Enum.__str__ change could be improved in What's New

[mdboom]

#30582 (in 3.11) introduced a change to Enum.__str__ so that it shows the enum name and the member name, whereas prior it only included the member name.

This could be a breaking change in some cases. As an example in real world code, the Azure CLI uses enums for the acceptable values of command-line flags, and then uses str on the enum to generate the string that is passed to the subprocess as arguments.

For example, this enum is used to define a "compute model":

class ComputeModelType(str, Enum):

    provisioned = "Provisioned"
    serverless = "Serverless"

And then it's used here to construct arguments for a subprocess:

self.cmd('sql db update -g {} --server {} --name {} --compute-model {}'
                 .format(resource_group, server, database_name, compute_model_serverless),

The changelog entry for this change just says:

IntEnum / IntFlag / StrEnum fixes: these now inherit from ReprEnum so the str() output now matches format() output, which is the data types’ (so both str(AnIntEnum.ONE) and format(AnIntEnum.ONE) is equal to '1').

This is unlikely to be easily findable for users running into this change, since StrEnum doesn't exist in Python 3.10, and it's a little obscure. I would suggest:

The default str() of an Enum has changed to include both Enum name and the member name (Enum.MEMBER), whereas before it was just the member name. To continue to use the old behavior, inherit from IntEnum or StrEnum.

Does this make sense? Are there other cases I'm missing? Happy to file a PR, but thought I would get some discussion first.

[ethanfurman]

I'm more than happy to have help crafting a better What's New entry. Keep in mind that the change you are talking about is format, not str -- str() output hasn't changed, format() has changed to match str().

[CAM-Gerlach]

Hey, I didn't see this before I first submitted #98295, but I'd noticed something similar, and I've revised that PR to further resolve this issue.